Robot Birds Could Help End Bird Strikes On Airplanes
Everyone knows about the Miracle on the Hudson, but you might not know that approximately one in every 2,000 flights experiences a bird strike. A vast majority of these don't end with a plane ditched in a river or a bird piercing a cockpit window, but this common problem is a constant threat to aviation safety.
In fact, between 1960 and 2014, bird strikes were responsible for the destruction of approximately 150 civil aircraft and the deaths of 271 people. So, it's no wonder that the Federal Aviation Administration places heavy emphasis on ensuring new aircraft and aircraft engines are well-prepared for this inevitability. In the US alone, the FAA estimates that there are 11,000 bird strikes annually, costing airlines and airports up to $957 million per year — and that's just hard costs, not factoring in lost time by passengers thanks to delays, diversions and cancellations due to damaged aircraft.
What if there was a way to reduce this problem, without resorting to killing off tens of thousands of birds at each airport? That's the challenge I just tackled for my PhD thesis in Computer Science at the University of Texas at Austin.
Because this is such a costly problem, most — but not all — airports do something to reduce bird strikes. Some methods are as simple as keeping the grass mowed, decreasing bird-friendly roosting areas and removing water sources. Other methods involve firing pyrotechnics, flying trained falcons or deploying Piper the border collie to scare birds away.
If birds become troublesome enough, airports sometimes resort to trapping and relocating them and/or killing them, like what's been done around LaGuardia. However, some airports have their hands tied by restrictions. Jackson Hole, Wyoming (JAC) has experienced dozens of bird strikes over the last couple of decades, many by the sage grouse, but can't destroy the bird's nearby nesting grounds because the sage grouse is a federally protected species.
Some of the most advanced current solutions simply involve adding pulsing lights to aircraft:
But even with these methods in place, bird strikes remain a costly issue. Airlines and airports would be very happy to find new, more effective solutions.
As you've probably seen, species of birds use either cluster or V-shaped flocking. V-shaped flocking is what it sounds like — birds flying in a V-shape, such as pelicans do at the beach. The behavior of V-flocks is primarily determined by the bird at the head of the flock.
Meanwhile, cluster flocking birds base their behavior on the behaviors of nearby birds. This type of flocking has evolutionary benefits — a flock is able to react to a danger even if only a small portion of the flock detects the danger. Starlings provide a beautiful example of cluster flocking:
Remote-Controlled Robot Birds
Some airports are trying to use robot birds to scare away birds. In May, we learned about how Edmonton (YEG) airport is using a human-piloted robot falcon to deter birds in a 13-week study over the summer. A robot bird is flown over farmland near the airport by a certified pilot and observer. The hope is that birds will avoid the airport when the robot falcon is flying because they'll believe a predator is patrolling the airport area.
One downside of this solution is that a pilot and observer must be actively involved at all times. Also, the remote-controlled bird merely scares birds away from the area, so it could accidentally scatter birds into an aircraft's flight path.
My solution seeks to avoid this by utilizing the birds' own flocking instincts to guide them away from the airport, rather than scaring them.
Influencing Flocks using Autonomous Robot Birds
Assume there is a cluster flock of birds heading toward an airport. To reduce the chance of a bird strike, it would be ideal to influence the flock to fly around the airport instead of through the airport. My research considered how autonomous — not human-controlled — robot birds should behave to join, influence and then leave a flock.
While a falcon-like robot bird is used at Edmonton to mimic a predator, my research builds on the assumption that the flock will see a robot bird as one of their own. Seems crazy, but researchers have actually found that birds see a robot bird as friendly if the robot bird has a similar silhouette and wing flap motion. There's even video of birds flocking with a robot:
Avian biologists have used observations to produce models of how different species flock. These models include both a "neighborhood model" and a "flocking model." The neighborhood model defines which surrounding birds influence each bird in a flock, and can look at the seven nearest birds or at all birds within a particular distance. The flocking model defines how each bird is influenced by its neighbors. Knowing the flocking model of a species is critical to determining how the robot birds should behave to influence a flock.
That's where my research kicks in: given limitations in number, computational power and battery life of the robot birds, how should the robot birds join a flock, behave within the flock and then leave the flock in order to optimally influence it away from danger?
Many methods were iteratively tested, analyzed, and refined in simulation using flocking models from biologists. Here's a video showing an algorithm in which I assume the robot birds (shown in pink) can hover as they strategically position themselves to influence the flock to turn:
In later research, my work assumes that the robot birds can't fly faster or slower than the flock, complicating how the birds join, influence and leave the flock while trying to keep the flock cohesive:
Finally, I implemented some of my behaviors on humanoid robots. These tests showed that my influencing behaviors work on a robot platform. In this video, the robot wearing the orange jersey influences the flock to travel around the field's center circle:
Now that this underlying research has been done, it's ready to be implemented on robot birds. These robot birds could then fly autonomously — in other words, without a pilot and observer.
To reduce bird strikes, the robot birds could leave charging stations around an airport, intercept each incoming flock, guide it around the airport, leave the flock and then return to their charging stations. There are regulatory, safety and power hurdles that would need to be considered, but the technology is feasible.
Dr. Peter Stone in the computer science department at The University of Texas at Austin was the advisor on my thesis, which is available on my web page.
What do you think about this approach? What types of bird reduction methods have you seen at airports?